Warm points keep trains running on time thanks to AI

February 15, 2026

Sensors ensure smooth rail operations in Scandinavia, even in very harsh winters

Points that cannot be set due to icing or large snow accumulations lead to massive delays, sometimes even train cancellations. Heating them is not enough, as William Liu from the Royal Institute of Technology (https://www.kth.se/en ) has discovered. In Sweden alone, although almost all points are heated, there are around 1,000 cancellations every year, affecting hundreds of thousands of passengers. Sensors and artificial intelligence (AI) are expected to provide the solution.

Better despite cost reduction

The sensors record the weather conditions at each individual switch and measure temperature, wind speed and other data (https://link.springer.com/chapter/10.1007/978-3-032-04774-8_41) that are responsible for the fate of the switches. From this, AI software that has been trained with countless weather and associated icing data derives the necessary heating temperature.

Currently, this is a constant eight degrees Celsius in Sweden, which is not sufficient in certain weather conditions. Conversely, the temperature can be lowered if there is no risk of icing despite the extreme cold. The actual heat loss also depends on factors other than the outside temperature, namely wind speed, humidity and the presence of snow. Wind in particular is a critical factor because it cools the metal more.

Computer model for forecasting

Liu (https://www.kth.se/profile/zhendong) has developed a computer model (https://www.kth.se/en/om/nyheter/centrala-nyheter/rail-switch-data-may-reduce-train-delays-in-cold-climates-1. 1456986) to predict how a switch will heat up under different weather conditions, including factors such as air temperature, wind and precipitation. The forecasts correspond to the data actually measured on a test switch. It has been shown that short periods during which the switch is heated to well over eight degrees are far more effective in critical situations than continuous heating.

Related Articles

Engineered disorder could make protective materials more resilient

Researchers at MIT and Northeastern University have developed a metamaterial with an irregular network structure that withstands impacts better than conventional ordered lattices. The findings could inform future helmet and body armour designs, although testing has so...

IAA MOBILITY 2027: Early bookings signal strong demand in Munich

Around ten months before the event opens, exhibitor registrations are well ahead of the comparable point in the previous cycle. The 2027 programme will put AI, autonomous driving, battery technology and industrial transformation in focus. By 12 November 2026, 166...

Netherlands: Operation Endgame targets three malware networks

International investigators have taken down infrastructure linked to Gootloader, CastleRAT and StealC. The Dutch police report action against 14 Gootloader servers, six CastleRAT servers and 80 StealC servers. Working with international partners, they also seized...

Share This